Every inventory decision you make—how much to order, when to order it, how much safety stock to carry—depends on one thing: what you expect to sell. Demand forecasting is how you make that expectation as accurate as possible.
Demand Forecasting Definition
Demand forecasting is the process of predicting future customer demand for your products. It uses historical sales data, market trends, and other factors to estimate how many units you'll sell over a given time period.
A good forecast isn't about being perfectly right—that's impossible. It's about being close enough that your inventory decisions work out more often than not.
Why Demand Forecasting Matters
It Drives Every Other Decision
Your demand forecast flows downstream into everything:
- Safety stock calculations depend on forecast accuracy
- Reorder points are based on expected daily demand
- Purchase orders are sized to meet forecasted needs
- Cash flow planning assumes certain inventory investments
- Warehouse capacity needs to handle expected volume
Get forecasting wrong, and all these downstream decisions are built on a shaky foundation.
The Cost of Getting It Wrong
Forecast too low: Stockouts, lost sales, disappointed customers, and scrambling for expedited shipments.
Forecast too high: Excess inventory, tied-up cash, storage costs, and potential markdowns or write-offs.
For CPG brands, forecasting errors compound quickly. A 20% miss on a hero SKU can mean tens of thousands in lost revenue or dead stock.
Types of Demand Forecasting
Short-Term Forecasting
Predicts demand over the next few weeks to months. Used for immediate operational decisions like replenishment orders and production scheduling. Focuses on precision at the SKU level.
Long-Term Forecasting
Looks 6-18 months ahead. Used for strategic decisions like capacity planning, budgeting, and new product development. Accepts less precision in exchange for directional guidance.
Qualitative Forecasting
Based on expert judgment, market research, and informed opinions. Useful when historical data is limited—like new product launches or entering new markets.
Quantitative Forecasting
Uses statistical methods applied to historical data. More objective and scalable, but requires sufficient historical data to be reliable.
Common Demand Forecasting Methods
Moving Average
Averages demand over a recent period (e.g., last 12 weeks) to smooth out fluctuations. Simple to understand and implement. Works well for stable products without strong trends or seasonality.
Exponential Smoothing
Weights recent data more heavily than older data. Responds faster to changes than simple moving averages. Good for products with gradual trends.
Seasonal Decomposition
Identifies and accounts for seasonal patterns in your data. Essential for products with predictable peaks and valleys—holiday items, summer products, back-to-school merchandise.
Regression Analysis
Identifies relationships between demand and causal factors like price, promotions, or economic indicators. Useful when you understand what drives your sales beyond just historical patterns.
The Demand Forecasting Process
Step 1: Gather Historical Data
Pull sales history at the level you want to forecast—by SKU, by channel, by location. The more granular your data, the more granular your forecast can be.
Step 2: Clean the Data
Remove anomalies that would skew your forecast. Stockout periods (where sales were limited by availability, not demand) are the biggest culprit. Also account for one-time events like a viral social media mention.
Step 3: Choose Your Method
Match the forecasting method to your product characteristics. Stable products can use simple methods. Seasonal or trendy products need more sophisticated approaches.
Step 4: Generate the Forecast
Apply your chosen method to produce demand estimates for future periods. Most businesses forecast weekly or monthly, depending on their planning cadence.
Step 5: Review and Adjust
Forecasts from statistical models are a starting point. Layer in your knowledge of upcoming promotions, new retail partnerships, market shifts, or other factors the model can't see.
Step 6: Measure and Improve
Track forecast accuracy over time. Learn where your forecasts are consistently off and adjust your approach.
Measuring Forecast Accuracy
The most common metric is Forecast Accuracy Percentage:
Forecast Accuracy = (Actual / Forecast) × 100
A result of 100% means you nailed it. 80% means you forecasted 20% more than actual demand. 120% means actual demand exceeded your forecast by 20%.
What's "good enough" depends on your product and industry. For CPG brands, 70-85% accuracy at the SKU level is often acceptable. Higher is always better, but perfection isn't realistic.
Factors That Make Forecasting Harder
New Products
No historical data means you're starting from scratch. Use analogous products, market research, and smaller initial orders until you build a demand history.
Promotional Activity
Promotions spike demand temporarily, then often create a post-promotion dip. Forecasting through promotions requires understanding their true lift, not just the spike.
Multiple Channels
DTC, Amazon, retail, and wholesale each have different demand patterns. Forecasting for multi-channel brands means managing multiple forecasts that roll up to a total.
External Factors
Economic conditions, competitor actions, weather, and social trends all influence demand in ways that historical data alone can't capture.
Key Takeaways
- Demand forecasting predicts future customer demand using historical data and market intelligence
- Forecasts drive all downstream inventory decisions—get this wrong and everything else suffers
- Match your forecasting method to your product characteristics
- Clean your data before forecasting—especially stockout periods
- Measure forecast accuracy and continuously improve your process
- Perfect forecasts don't exist; aim for "good enough" and build in safety buffers
Frequently Asked Questions
Q: What is demand forecasting?
Demand forecasting is the process of predicting how much of your products customers will buy over a future time period. It uses historical sales data, market trends, and other factors to estimate future demand.
Q: Why is demand forecasting important?
Every inventory decision depends on your demand forecast—how much to order, when to order, how much safety stock to carry. Poor forecasts lead to stockouts or excess inventory, both of which hurt your business.
Q: What's a good forecast accuracy?
For CPG brands at the SKU level, 70-85% accuracy is typically acceptable. Higher is better, but 100% accuracy isn't realistic. Build safety stock to buffer against forecast error.
Q: How do you forecast demand for a new product?
Use analogous products (similar items you've sold before), market research, pre-order data, and expert judgment. Start with conservative quantities and adjust as real sales data comes in.
Q: How often should I update my forecast?
Most brands update forecasts monthly or weekly, depending on their planning cadence. Always update when you have significant new information—like landing a major retail account or seeing unexpected demand shifts.